Snorkel AI's $3.5B bet on training data

Snorkel AI has raised a $350 million Series E round at a $3.5 billion valuation, the company announced. The figure nearly triples the $1.3 billion valuation Snorkel carried when it raised a $100 million Series D 17 months ago, according to TechCrunch.

The round was led by Insight Partners and S32, with existing investors Addition, Lightspeed, Greylock, GV, and Wells Fargo also participating. TechCrunch reported the funding on September 22, 2026.

A valuation that tripled in 17 months

Snorkel's jump from a $1.3 billion valuation to $3.5 billion came in a span of 17 months, according to the report. That interval is shorter than the typical gap between late-stage rounds for many venture-backed companies, though the source does not characterize it as unusual.

The Series E was led by Insight Partners and S32. Existing backers Addition, Lightspeed, Greylock, GV, and Wells Fargo also participated, per TechCrunch's reporting.

The seven-year-old startup has now raised a total that includes the $350 million Series E. The source does not disclose prior cumulative funding beyond the $100 million Series D.

From labeling tool to data-as-a-service

TechCrunch reported that Snorkel originally provided software for data-labeling automation. Last year, the company shifted to providing customers with completed datasets, an offering it calls data-as-a-service, according to the report.

Rather than operating purely as a human expert marketplace, Snorkel relies on a hybrid approach, using its software and models to generate data synthetically alongside subject matter experts, TechCrunch reported.

The company says its current annualized revenue run rate stands at $375 million, an eighteenfold increase over the last 12 months. TechCrunch attributes that growth to AI labs' appetite for high-end training data.

Revenue growth that dwarfs the round size

The $375 million annualized run rate is more than the $350 million raised in the Series E. That figure represents an eighteenfold increase over 12 months, according to the company.

TechCrunch noted that other data companies positioning themselves as AI data labs have seen similar growth. Mercor's gross annualized revenue has climbed to $2 billion, Handshake hit the $1 billion milestone earlier this year, and TechCrunch reported that Micro1 has scaled to $500 million.

The source points out that since these companies pay out roughly 60% to 70% of their top-line income directly to the domain specialists doing the work, their actual net annual revenue is substantially lower than those headline gross figures. Snorkel says it sells reinforcement learning environments and complete datasets rather than human labor, so payments to its human experts are accounted for in its cost of goods sold rather than headline-generating annualized revenue numbers, according to the company.

• $350 million — Series E round size
• $3.5 billion — post-money valuation
• $1.3 billion — prior valuation 17 months ago
• $100 million — Series D round size
• $375 million — current annualized revenue run rate
• 18x — revenue growth over the last 12 months
• 60% to 70% — share of top-line income paid to domain specialists at comparable firms

What the data-as-a-service model does

Snorkel's offering combines synthetic data generation with human subject matter experts. The company moved away from being purely a labeling-software vendor, according to the report.

Instead of selling tools, Snorkel now sells completed datasets. It also sells reinforcement learning environments, which are used to train and evaluate AI models.

TechCrunch reported that Snorkel launched commercially in 2019 following four years of research by co-founder and CEO Alex Ratner and his team at a Stanford AI lab.

The economics of AI data work

The 60% to 70% payout figure for domain specialists is central to understanding the economics of the AI data-lab category. For companies structured as human expert marketplaces, most gross revenue flows to the workers doing the labeling and domain work.

Snorkel's structure differs, according to the company. Because it sells datasets and reinforcement learning environments rather than human labor, payments to experts are classified as cost of goods sold rather than being counted in the headline annualized revenue figure.

That accounting difference means Snorkel's $375 million run rate is not directly comparable to the gross figures reported by Mercor, Handshake, or Micro1, per the source.

Competitors riding the same wave

Mercor's gross annualized revenue has climbed to $2 billion, according to TechCrunch. Handshake hit the $1 billion milestone earlier this year. Micro1 has scaled to $500 million, TechCrunch reported.

These figures are gross annualized revenue, not net. The source notes that roughly 60% to 70% of that top-line income goes directly to the domain specialists doing the work at those companies.

Snorkel positions itself differently by selling completed datasets and reinforcement learning environments rather than human labor by the hour, according to the company.

The Stanford research roots

Snorkel launched commercially in 2019, following four years of research by co-founder and CEO Alex Ratner and his team at a Stanford AI lab, according to the report.

That research origin predates the current boom in AI training data demand. The company was founded seven years ago, making it roughly seven years old as of the report.

Ratner remains co-founder and CEO, as named in the source.

What the round signals for AI infrastructure

The Series E comes 17 months after the Series D. The valuation step-up from $1.3 billion to $3.5 billion reflects both the company's revenue growth and the broader demand environment for AI training data.

Insight Partners and S32 led the round, with Addition, Lightspeed, Greylock, GV, and Wells Fargo participating. The participation of existing investors suggests continued conviction in the company's direction.

The source does not disclose specific plans for the new funding, nor does it detail how the capital will be deployed.

Why this matters for the AI stack

The growth figures across Snorkel, Mercor, Handshake, and Micro1 point to how AI companies are sourcing their training material. If these numbers hold, the market for curated, task-specific datasets and reinforcement learning environments appears to be expanding rapidly.

For businesses building on AI, the availability of high-quality training data affects what models can do. The companies that can deliver that data at scale — and do so profitably — could be positioned to capture a growing share of AI development budgets.

The accounting distinction between gross annualized revenue and net revenue matters for anyone evaluating this sector. Headline figures at human-marketplace companies overstate net revenue because most of the money flows to specialists. Snorkel's structure, by the company's account, avoids that specific distortion.

For the industry, the question is whether the current growth rates are sustainable. AI labs' demand for training data is tied to their own development cycles and funding availability. If model development slows or shifts toward techniques that require less external data, the demand curve could flatten. For now, the numbers reported by these companies describe a market expanding quickly, with investors willing to fund the infrastructure layer that feeds it.